An economic analysis of induction of labour and expectant monitoring in women with gestational hypertension or pre-eclampsia at term (HYPITAT trial)

Authors


SMC Vijgen, Department of Obstetrics and Gynaecology, Academic Medical Centre Amsterdam, PO Box 22660, 1100 DD Amsterdam, the Netherlands. Email s.m.vijgen@amc.uva.nl

Abstract

Please cite this paper as: Vijgen S, Koopmans C, Opmeer B, Groen H, Bijlenga D, Aarnoudse J, Bekedam D, van den Berg P, de Boer K, Burggraaff J, Bloemenkamp K, Drogtrop A, Franx A, de Groot C, Huisjes A, Kwee A, van Loon A, Lub A, Papatsonis D, van der Post J, Roumen F, Scheepers H, Stigter R, Willekes C, Mol B, Van Pampus M for the HYPITAT study group. An economic analysis of induction of labour and expectant monitoring in women with gestational hypertension or pre-eclampsia at term (HYPITAT trial). BJOG 2010;117:1577–1585.

Objective  To assess the economic consequences of labour induction compared with expectant monitoring in women with gestational hypertension or pre-eclampsia at term.

Design  An economic analysis alongside the Hypertension and Pre-eclampsia Intervention Trial At Term (HYPITAT).

Setting  Obstetric departments of six university and 32 teaching and district hospitals in the Netherlands.

Population  Women diagnosed with gestational hypertension or pre-eclampsia between 36+0 and 41+0 weeks of gestation, randomly allocated to either induction of labour or expectant monitoring.

Methods  A trial-based cost-effectiveness analysis was performed from a societal perspective during a 1-year time horizon.

Main outcome measures  One-year costs were estimated and health outcomes were expressed as the prevalence of poor maternal outcome defined as either maternal complications or progression to severe disease.

Results  The average costs of induction of labour (n = 377) were €7077 versus €7908 for expectant monitoring (n = 379), with an average difference of −€831 (95% CI −€1561 to −€144). This 11% difference predominantly originated from the antepartum period: per woman costs were €1259 for induction versus €2700 for expectant monitoring. During delivery, more costs were generated following induction (€2190) compared with expectant monitoring (€1210). No substantial differences were found in the postpartum, follow-up and for non-medical costs.

Conclusion  In women with gestational hypertension or mild pre-eclampsia at term, induction of labour is less costly than expectant monitoring because of differences in resource use in the antepartum period. As the trial already demonstrated that induction of labour results in less progression to severe disease without resulting in a higher caesarean section rate, both clinical and economic consequences are in favour of induction of labour in these women.

Trial registration  The trial has been registered in the clinical trial register as ISRCTN08132825.

Introduction

Six to eight percent of all pregnancies are complicated by gestational hypertension or pre-eclampsia.1,2 Although outcome in most cases is good, hypertensive diseases remain a major cause of morbidity and mortality for both mother and child.3 Moreover, the care for women with hypertensive disease in pregnancy imposes a substantial economic burden.

Most hypertensive diseases occur at or near term.4 Because evidence on the choice between induction of labour and expectant monitoring for women with gestational hypertension or mild pre-eclampsia at term is lacking, we recently performed a randomised clinical trial on that subject:5,6 the Hypertension and Pre-eclampsia Intervention Trial At Term (HYPITAT, number ISRCTN08132825). Induction of labour resulted in significantly fewer women with progression to severe disease and the caesarean section rate was lower, albeit not significantly. Apart from these clinical outcomes, knowledge on the cost is also of importance, to decide whether induction should be applied or not. At present, evidence on costs and cost-effectiveness of management of women with gestational hypertension or pre-eclampsia at term is limited.

This study reports the results of the economic evaluation that we performed alongside the HYPITAT trial. We performed a cost-effectiveness analysis comparing induction of labour with expectant monitoring in pregnancies at high risk because of hypertensive disorders beyond 36 weeks of gestation.

Methods

Trial design

Full details of the HYPITAT trial were reported previously.5 The trial was approved by the Institutional Review Board of the University of Leiden (p04.210) and had local approval from the boards of the other participating hospitals. The trial has been registered in the clinical trial register as ISRCTN08132825.

In short, the study was a multicentre randomised controlled clinical trial conducted between October 2005 and March 2008 in the obstetric departments of six university and 32 teaching and district hospitals in the Netherlands. The participating hospitals are listed in the Supporting Information, Appendix S1. Women diagnosed with gestational hypertension or pre-eclampsia beyond 36 weeks of gestation were allocated to either induction of labour or expectant monitoring.

In the induction group, labour was induced within 24 hours after randomisation. If the cervix was ripe, labour was induced with amniotomy and if labour did not start within 1 hour augmentation with oxytocin was used. If the cervix was judged to be ‘unripe’, cervical ripening was stimulated using intracervical or intravaginal prostaglandins, according to the local protocol.

In the expectant group, women were monitored following the local protocol until the onset of spontaneous delivery. Maternal monitoring consisted of frequent blood pressure measurements and laboratory tests. Fetal monitoring consisted of assessment of fetal movements as reported by the mother, as well as electronic fetal heart rate monitoring and ultrasound examination. Intervention was only recommended in the case of one or more of the following conditions: diastolic blood pressure ≥110 mmHg, systolic blood pressure ≥170 mmHg, proteinuria ≥5 g/24 hours, eclampsia, HELLP syndrome (haemolysis, elevated liver enzymes and low platelet count), suspected fetal distress, prelabour rupture of membranes lasting >48 hours, meconium-stained amniotic fluid or gestation beyond 41 weeks.

Women who did not give informed consent for randomisation but who gave authorisation for the use of their medical records were treated according to one of the two protocols at the discretion of the attending obstetrician. Data of these women were analysed to compare them with the randomised data.

The primary outcome in this trial was a composite measure of poor maternal outcome, defined as maternal mortality, maternal morbidity (eclampsia, HELLP syndrome, pulmonary oedema, thromboembolic disease or placental abruption), progression to severe disease (at least one measurement during antenatal or postpartum period of diastolic blood pressure ≥110 mmHg, systolic blood pressure ≥170 mmHg and proteinuria ≥5 g/24 hours) or major postpartum haemorrhage6. Secondary outcomes were: method of delivery, neonatal mortality and neonatal morbidity. For neonatal morbidity we used a composite outcome consisting of a 5-minute Apgar score < 7, umbilical artery pH <7.05, or admission to a neonatal intensive care unit.

Analysis of the clinical endpoints showed fewer women with severe disease in the induction group (31 versus 44%, relative risk 0.71; 95% CI 0.59–0.86). No maternal or neonatal death or eclampsia occurred in either group. There was no difference in neonatal morbidity rate (6 versus 8%, relative risk 0.75; 95% CI 0.45–1.26) and caesarean section rate (14 versus 19%, relative risk 0.75; 95% CI 0.55–1.04).6

Economic evaluation

An economic analysis was performed alongside the trial. As one strategy was found to be more effective the economic evaluation was set up as a cost-effectiveness analysis.7,8 All unit costs were expressed in 2007 Euros using the consumer pricing index.9

We used a societal perspective, which means that we included both medical and nonmedical costs to examine the economic impact of both strategies on the whole society. We compared costs and effects from the moment of randomisation to 1 year postpartum. Thereby, we differentiated different cost categories (direct medical, nonmedical and indirect costs) and provided details on use of healthcare resources. Discounting the costs was unnecessary because all costs occurred within 1 year.

Measuring resource use

Resource use was documented by extending the Case Record Form with specific items on healthcare use and by administering additional questionnaires. In the Case Record Form the following resource items were collected: maternal and neonatal admissions, method of delivery, induction method, hours in labour room and/or operating theatre, outpatient visits, medication, maternal laboratory tests, fetal monitoring, third-stage delivery activities and neonatal monitoring.

Maternal admissions were differentiated into three phases: the antenatal, the delivery and the postpartum phases. For each admission, hospital stay was differentiated according to the level of care: intensive care, medium care, maternal ward or home care, because different levels have different costs.

The time in the labour room was calculated as the time from admission to labour room to time of birth plus 1 hour extra for recovery care. Assuming that induction of labour takes place inside the labour room, it will be expected that the mean number of hours in the labour room are higher in the induction group because of the time needed for induction. If a caesarean section was performed, hours in the operating theatre were also calculated.

For each neonatal admission, hospital stay was differentiated according to the level of care as well: intensive care, high care, medium care, medium care on maternal ward or maternal ward, because different levels have different costs. Duration of neonatal admission was calculated as the number of days between birth and hospital discharge. For neonatal admission to the maternal ward no extra costs were generated because it was assumed that these costs were already included in those for the mother.

The long-term use of healthcare from hospital discharge to 1 year postpartum was collected using an additional questionnaire sent to a consecutive subsample of 99 randomised women and 48 nonrandomised women 1 year after their childbirth. The questionnaire documented visits to healthcare providers and hospital admissions for their child and for themselves and medication use. In addition to these medical costs, sick leave from work (indirect nonmedical costs), modes of travelling to hospital and the use of informal care given by partner or family (direct nonmedical costs) were assessed.

Unit costs

Different methods and sources were used to estimate unit costs as valuations for documented volumes of resource use (Table 1). For maternal and neonatal admissions, third-stage delivery and neonatal monitoring, unit cost estimates were available from the financial departments of one participating academic hospital and one participating general hospital.

Table 1.   Cost-analyses: units of resource use, unit costs, valuation method and volume source
 UnitUnit costValuation method (source)Volume source
  1. AQ, additional questionnaire; CRF, Case Record Form.

  2. *The mean of the unit cost for an academic hospital and for a general hospital is presented.

  3. **The mean of several methods/medications is presented.

Medical costs
Admission mother*
  hospital stay—wardday346Top-down calculationCRF
  hospital stay—medium careday526Top-down calculationCRF
  hospital-stay–intensive careday1679Top-down calculationCRF
Admission child*
  hospital stay—medium careday526Top-down calculationCRF
  hospital stay—high careday1409Top-down calculationCRF
  hospital-stay—NICUday1459Top-down calculationCRF
Specialist carehour69Dutch costing guidelines (11)CRF/AQ
Outpatient visit*visit82Top-down calculationCRF/AQ
Psychologisthour34Dutch costing guidelines (11)AQ
Midwifehour34Dutch costing guidelines (11)AQ
General practitionervisit21Dutch costing guidelines (11)AQ
Paramedicalvisit25Dutch costing guidelines (11)AQ
Home carehour32Dutch costing guidelines (11)AQ
Day careday242Dutch costing guidelines (11)CRF
Induction methods**gift15Pharmacotherapeutic website (12)CRF
Antihypertensive medication and antibiotics**dose per day7Pharmacotherapeutic website (12)CRF
Analgesics during labour**procedure161Top-down calculationCRF
Neonatal monitoring**procedure90Top-down calculationCRF
Operation room*hour140Bottom-up calculationCRF
Labour room*hour82Bottom-up calculationCRF
Non-medical costs
Travel costs—carkm0.17Dutch costing guidelines (11)AQ
Travel costs—public transportkm0.17Dutch costing guidelines (11)AQ
Informal carehour8.78Dutch costing guidelines (11)AQ
Productivity losshour26Dutch costing guidelines (11)AQ

For use of the labour room and the operating theatre, unit costs were calculated per hour, using a bottom-up approach, in which all personnel, use of materials and overheads, calculated as a square metre price, were integrated. Costs per type of delivery were then calculated by multiplying time in labour room by the price for 1 hour in the labour room and by counting up time in the operating theatre multiplied by price per hour in operating theatre.

For some cost units (outpatient visit, specialist care, general practitioner visit, paramedical and home care, travel costs, informal care and productivity loss) national standardised prices were used, and for laboratory testing, published tariffs were used.10,11 Medication prices were estimated using the Pharmacotherapeutic Compass.12

The value of productivity loss was calculated using the friction cost method from age- and sex-stratified data of the Dutch population.10,11 This method assumes that workers that are withdrawn from work through ill health will be replaced after some adaptation period—the ‘friction period’. Consequently, costs from an individual woman’s production loss are limited to a period of 10 weeks.

Analyses

Group differences in resource use were tested using the nonparametric Mann–Whitney U test, as resources generally have a skewed distribution. Resource use per woman was multiplied by unit costs, and total costs per woman were calculated. Mean costs and median costs per woman were estimated, and mean cost differences between study groups were calculated. The 95% CI around the difference in mean costs, and incremental cost-effectiveness ratios (ICERs) were determined by bootstrapping. An ICER is the ratio of cost differences versus the effectiveness differences between two interventions. Bootstrap methods are based on generating multiple replications of the statistic of interest by sampling with replacement from the original data.7,8 Analyses were by intention-to-treat.

Thirteen univariate sensitivity analyses were performed to explore the impact of different assumptions and alternative unit-cost estimates on the results of the costs analysis. Several assumptions were made in estimating labour and operating theatre costs by using a bottom-up method, such as time spent in labour room and/or operating theatre by obstetricians and gynaecologists. In the first sensitivity analysis some variations of these assumptions were made to discover their impact on the final results (model 1). We examined several other ways of estimating the delivery costs using a top-down method (model 2) and a combination of both methods (model 3). Because most cost differences were expected antepartum because of longer maternal hospital stays in the expectant group, we wanted to find out the impact of lower valuation of the antepartum admissions by assuming several other monitoring strategies: medium care instead of intensive care admissions (model 4), daycare instead of inpatient care (model 5), outpatient visits plus cardiotocograms instead of inpatient care (model 6) and home care instead of inpatient care (model 7). In our base-case analysis we included no costs for the neonatal ward admissions because we assumed that this was covered by the maternal ward admissions. In model 8 we priced neonatal ward admissions to check their impact. In the base-case analysis we separated all admissions into four different phases. In a sensitivity analysis we examined the impact of no separation into phases (model 9). In another sensitivity analysis (model 10) operating theatre costs were calculated by assuming 1 hour operating theatre time per woman instead of the actual measured hours in the original study. Finally, the impact of using lower or higher unit costs during all the phases was studied in models 11, 12 and 13.

A subgroup analysis was performed to assess the consistency of the cost-effectiveness effect in women with gestational hypertension and women with mild pre-eclampsia.

Statistical, economic and simulation analyses were performed using spss software (version 16.0, Chicago, IL, USA) and Microsoft Excel.

Results

Resource use

For the cost analysis we used the data from all 756 randomised women and all 397 nonrandomised women. Of the randomised women, 377 were assigned to the induction group and 379 to the expectant monitoring group. Among the 397 nonrandomised women, labour was immediately induced in 73, and 324 women were initially managed expectantly. Additional data on resource use during follow-up acquired by questionnaires were available for 55 randomised women, 32 allocated to induction and 23 allocated to expectant management. In the nonrandomised group 41 women returned a questionnaire, of whom six had been allocated to induction and 35 had been monitored expectantly.

Average volumes of resource use, total costs in each study group, and average costs per woman are presented in Supporting Information Tables S1 and S2. Most remarkable differences in resource use between groups were generated during the antenatal and delivery periods. There were more outpatient visits in the expectant group (9% in induction group versus 66% in expectant group, P < 0.001) and a longer antenatal stay in hospital (3 days in the induction group versus 6 days in the expectant group, P < 0.001). Stay in the labour room or operating theatre was longer for induced women (21 versus 10 hours, P < 0.001). Until 1 year after childbirth women in the expectant group stayed a little longer in hospital than the women in the induction group (3.5 days induction versus 4.0 days expectant, P < 0.001). The number of hospital days for the children was comparable between both groups (4.2 days induction and 4.3 days expectant, P = 0.014). The average duration of sick leave after the 3 months of permitted maternal leave, which is common practice in the Netherlands, until 1 year follow up was comparable (3.0 versus 2.3 days, P = 0.39).

Costs

A summary of mean and median costs per woman is presented in Table 2. In the antepartum period costs per woman appeared to be higher in the expectant monitoring group because of longer maternal stays (difference: −€1441). During delivery the costs in the induction group were higher than in the expectant monitoring group (difference: €980). This is because of longer stays in the labour room associated with the induction procedure. Until 1 year postpartum, women in the expectant monitoring group generated slightly more costs than women in the induction group (difference: −€398), because of longer maternal stays, longer neonatal intensive care stays and more specialist visits. There were no substantial differences in nonmedical costs. Overall, mean costs per woman were €7077 (95% CI 2.326–19.726) for induction and €7908 (95% CI 2.561–27.037) for expectant monitoring (difference −€831; 95% CI −1.561 to −144).

Table 2.   Comparison of costs between randomised induction of labour and expectant monitoring
 Induction (n = 377)Expectant management (= 379)
MeanMedian (IQR)MeanMedian (IQR)
  1. IQR, Interquartile range.

  2. *No median and interquartile percentiles are presented, because these are extrapolated data.

  3. **Induction minus expectant management.

  4. ***Nonparametric confidence interval based on 1000 bootstrap replications.

Maternal admission993574 (0–1435)21361215 (0–2835)
Cardiotocograms and ultrasound168116 (87–203)287232 (145–377)
Outpatient visits120 (0–0)13159 (0–280)
Assessments and medication8163 (31–116)137116 (62–187)
Laboratory tests54 (2–6)98 (4–12)
Total antepartum1259822 (234–1786)27001817 (631–3466)
Admission because of labour319287 (0–574)285287 (0–405)
Induction material6445 (1–90)200 (0–1)
Medication during labour730 (0–161)640 (0–161)
Mode of delivery1734960 (584–1691)841640 (324–1,053)
Total delivery21901292 (815–2446)1210927 (567–1546)
Maternal admission1098861 (574–1435)1291861 (574–1,435)
Neonatal admission9670 (0–0)10120 (0–0)
Paediatrician and monitoring420 (0–69)4269 (0–69)
Total postpartum21051148 (810–2094)23451148 (643–2,417)
Primary and specialist care486458 (458–559)772957 (957–1452)
Maternal admission0114103 (103–146)
Neonatal admission240218 (218–218)0
Total follow up726677 (677–867)886759 (759–1,066)
Travel costs*22 37 
Informal care*703 686 
Productivity loss*72 44 
Total nonmedical care797 767 
Total cost70775530 (4142–7949)79086235 (4508–9331)
Differential mean cost** (95% CI)***−831 (−1561 to −144)

Table 3 shows mean and median costs per woman in the nonrandomised groups. Because only six women from the nonrandomised induced group returned the follow-up questionnaire, estimates of costs generated during follow up and nonmedical costs for this group are not reliable. If follow-up and nonmedical costs are excluded then total mean costs between randomised and nonrandomised induced women are comparable (€5554 versus €5670). Mean costs per nonrandomised woman in the expectant management group were lower than in the randomised expectant monitoring women in all phases except delivery.

Table 3.   Costs per woman in nonrandomised groups
 NR induction (n = 73)NR expectant management (n = 324)
MeanMedian (IQR)MeanMedian (IQR)
  1. IQR, interquartile range; NR, not randomised.

  2. *No median and interquartile percentiles are presented because these are extrapolated data.

Maternal admission1465574 (0–1325)1378405 (0–1722)
Cardiotocogramss and ultrasound187145 (116–218)233203 (145–290)
Outpatient visits780 (0–59)142104 (0–208)
Assessments and medication8154 (31–124)11793 (41–171)
Laboratory tests76 (4–8)76 (4–10)
Total antepartum1818951 (290–1550)18821113 (414–2231)
Admission because of labour309287 (0–574)257287 (0–405)
Induction material5645 (1–90)270 (0–45)
Medication during labour710 (0–161)600 (0–161)
Mode of delivery1358960 (589–1529)983653 (385–1234)
Total delivery17951267 (819–2364)1327945 (560–1711)
Maternal admission1309861 (810–1435)1110861 (574–1435)
Neonatal admission7190 (0–0)6650 (0–0)
Paediatrician and monitoring300 (0–69)4569 (0–69)
Total postpartum20571148 (836–2091)1818930 (724–1843)
Primary and specialist care491471 (471–530)579526 (526–709)
Maternal admission00
Neonatal admission0179161 (161–227)
Total follow up491471 (471–530)758687 (687–933)
Travel costs*35 25 
Informal care*220 403 
Productivity loss*0 112 
Total nonmedical care255 540 
Total cost64164924 (3443–7151)63255037 (3441–7695)

The subgroup analysis showed mean costs per woman with mild pre-eclampsia of €7870 (= 123, median €6347, interquartile range [IQR] €4826 to €9429) in the induction group and €10,387 (n = 123, median €8069, IQR €6019 to €12,792) in the expectant management group. Mean costs per woman with only gestational hypertension were €6679 (n = 244, median €5115, IQR €3789 to €7647) in the induction group and €6682 (n = 252, median €5377, IQR €4100 to €7876) in the expectant management group.

Cost-effectiveness

With an estimated difference in progression to severe disease between the two strategies of 13% in favour of induction of labour, and a mean difference in costs per woman of €831 also in favour of induction of labour, induction of labour was the dominant strategy.

We also assessed uncertainty in the estimated ICERs (bootstrap analysis), depicted in a cost-effectiveness plane. Figure 1 shows that with high certainty induction is a cost-saving strategy compared with expectant monitoring in pregnant women with gestational hypertension or mild pre-eclampsia. In the upper right quadrant a strategy is considered cost-effective if the ICER is located below the line reflecting willingness-to-pay (how much one is willing to pay for one unit gain in health outcome). A strategy will always be cost-effective, irrespective of willingness-to-pay, if all bootstrap estimates of incremental cost-effectiveness ratios are located in the lower right quadrant.

Figure 1.

 Cost-effectiveness plane. Additional costs: difference in average costs per patient between the induction group and the expectant group (in 2007 Euros). Additional effects: difference in proportion of maternal complications between the induction group and the expectant group.

Sensitivity analyses

In Table 4 the results of the sensitivity analyses are shown. If the labour room and operating theatre costs are increased from €80 to €115 and from €140 to €224 per hour, respectively, this increases mean costs in both groups but decreases the difference to −€470, and expectant monitoring remains the most expensive strategy (model 1). Top-down calculation (model 2) or a combination of bottom-up and top-down calculation (model 3) of the delivery costs resulted in higher mean costs per woman in the expectant group and lower mean costs per induced woman. The antepartum intensive care admissions were valued for medium care prices, to find out if it impacted upon the total costs per group. The mean costs per woman in the expectant group decreased a little. Because costs for the induced woman remained similar, the mean cost difference fell to −€438 (model 4). Replacing antepartum inpatient care by other monitoring strategies (model 5, 6 and 7) decreased the mean costs per woman in both groups, but expectant management remains the most expensive strategy. Including neonatal ward admissions (model 8), summarising all admissions without separate phases (model 9) or standardising operating theatre time (model 10) increases mean costs per woman in both groups, but the differences remain similar. We also estimated costs by using higher and lower unit prices for hospital stay and outpatient visits. If we only use the unit costs from the academic centre, the mean costs per woman in both groups increase and the difference increases to −€1496 (model 11). If we only use the unit costs from the general hospital, the mean costs per woman decrease, but the differences between groups remain the same (model 12). If we estimate costs associated with inpatient admissions by using the Dutch national standardised unit costs the difference between groups increases (model 13).

Table 4.   Sensitivity analyses results
ModelDescriptionInductionExpectant managementDifference
 Base-case scenario€ 7077€ 7908−€ 831
 1Higher labour (€115) and operating (€224) theatre costs€ 7774€ 8244€ 470
 2Top-down calculation of delivery costs€ 6660€ 8456€ 1796
 3Bottom-up and top-down calculation of delivery costs (combining registered days of induction and top-down unit prices for delivery)€ 6897€ 8551€ 1654
 4Antepartum intensive care admissions priced as medium care€ 7007€ 7445€ 438
 5Value antepartum admissions by daycare prices€ 6817€ 7460€ 643
 6Replace antepartum admissions by outpatient visits and cardiotocogramss€ 6438€ 6804€ 366
 7Replace antepartum admissions by home care€ 6854€ 7523€ 669
 8Value neonatal ward admissions€ 7579€ 8309€ 730
 9Summarise all admissions without separate phases€ 8189€ 8899€ 710
10Standardise time in operation room to 1 hour (plus 1 hour recovery time)€ 7093€ 7939€ 846
11Value admissions by using academic unit prices only€ 8169€ 9665€ 1496
12Value admissions by using general unit prices only€ 6591€ 7359€ 768
13Value admissions by using Dutch standard prices (11)€ 7555€ 8774€ 1219

Discussion

This study assessed the economic consequences of induction of labour or an expectant monitoring strategy in pregnant women with gestational hypertension or mild pre-eclampsia at term, from a societal point of view. This economic analysis was performed alongside the HYPITAT trial.6 To our knowledge it is the first economic evaluation that prospectively compared these strategies.

Our analyses show that the mean costs per woman generated by induction of labour were €831 (95% CI −1.561 to −144) lower than those for expectant monitoring. Unsurprisingly, the difference in costs predominantly originated in the antepartum period, because of longer hospital stays before childbirth for women in the expectant group. On the other hand, during delivery more costs were generated following induction because of the longer time in the labour room because of induction itself. In our base-case analysis we assumed that induction was performed in the labour room and we estimated delivery costs by using number of hours in labour room and unit costs for 1 hour in the labour room. In sensitivity analyses we examined the consequences if delivery costs were estimated using a top-down calculation, i.e. using unit costs per method of delivery. As a result the cost difference between both strategies doubled. This means that if induction time is not included separately in the cost analyses (it is assumed that these costs are already included in the costs of hospital admission), the cost difference during delivery in favour of expectant monitoring disappears and expectant monitoring even becomes more expensive. However, in our opinion top-down calculation for estimating delivery costs is not an appropriate method, because it underestimates costs of induction and the time spent on induction itself.

In the period from childbirth until 1 year postpartum medical costs appeared to be €398 higher for women in the expectant monitoring group, mostly because of longer maternal and neonatal stays in the hospital and more postpartum specialist visits. The clinical results of the HYPITAT trial already indicated that induction of labour results in less progression to severe disease and that it has a lower caesarean section rate. The higher postpartum costs for women in the expectant group are a consequence of these findings. During delivery, expectant monitoring remained less costly despite more caesarean sections in this group. Higher caesarean section costs for induced women are caused by the longer stay in the labour room and operating theatre. No substantial differences were found in the nonmedical costs. These costs and the above-mentioned follow-up costs were estimated in a subgroup of the total trial population using an additional questionnaire. Because of the small number of participants, this results in broad confidence intervals around the follow-up and non-medical costs, and hence greater uncertainty around these estimates. On the other hand, we were able to estimate the impact of both strategies on indirect costs and for a longer time horizon than the trial duration, which improves the relevance from a societal perspective.

In the Netherlands, women are permitted maternal leave for 3 months after childbirth. In estimating the yearly costs of productivity loss we only included the remaining 9 months after maternal leave. Therefore, before generalising nonmedical costs to other countries, the local arrangements for maternal leave need to be considered.

The prospective design of the trial, the large number and diversity of participating hospitals and the well-organised structure of randomisation and data collection within the Dutch Obstetric Consortium13 are likely to extend both the internal and external validity of our results.

Overall, costs per randomised woman in the expectant group were higher than the costs per nonrandomised woman, except for the delivery phase. From the clinical data it can be observed that randomised women did not have many more bad maternal outcomes after expectant management than nonrandomised women (44 versus 38%). For that reason the randomised women might be admitted more often and for longer to hospital before and after childbirth and this might explain the higher costs per randomised woman in these groups. Children in the randomised expectant group stayed longer in MC/HC than the nonrandomised children. This is probably not a trial effect (better monitoring in the randomised group), because this effect was not seen within the induced groups.

According to the clinical results of the HYPITAT trial induction of labour should be advised in women with gestational hypertension and a diastolic blood pressure ≥ 95 mmHg or mild pre-eclampsia at a gestational age beyond 37 weeks.6 The results as described in this economic evaluation study indicate that this strategy is also associated with lower average costs per woman.

In summary, induction of labour is found to be a less costly and more effective strategy compared with expectant management in women with gestational hypertension or pre-eclampsia beyond 36 weeks of gestation.

Disclosure of interests

None declared.

Contribution to authorship

All authors have seen and approved the final version of the manuscript.

Details of ethical approval

The HYPITAT trial was approved by the Institutional Review Board of the University of Leiden (p04.210) and had local approval from the Boards of the other participating hospitals. The trial has been registered in the clinical trial register as ISRCTN08132825.

Funding

Financial support for this research was provided by ZonMw, the Hague, the Netherlands, grant number 945-06-553.

Acknowledgements

The authors would like to thank all members of the Dutch obstetric consortium (http://www.studies-obsgyn.nl/). We would especially like to thank the research nurses and midwives, who among others made this study possible. We would especially like to thank Drs ZA van Dijk and Mrs M Kruijt for their administrative support during the trial. We thank ZonMw for the financial support (grant number 945-06-553). See Supporting Information Appendix S1.

Ancillary